The XQL problem

[Ann Tecklenburg <at@ingenuity-sw.com>]

For those of you who are wondering about the state of my efforts vs. the
openXQL project:  I will say that
in trying to write a truly compliant XQL parser, I have already hit the
problems below.  The OpenXQL code goes around problems by omitting
anything that causes problems.  Ultimately this is unacceptable.  Antlr
itself is part of the problem since it cannot do unicode.    Both
SableCC and ANTLR die on the "lack of compositionality" described below.

Following is an excerpt from the XQL mailing list which describes just a
few of the problems with the proposed *final* version of the XQL
spec.    That XQL will become a "big thing" seems a fact.  That you can
build a compliant implementation may be something else.  The fact that
the the authors of the XQL proposal are engaging in a dialog in the
mailing list is great---but that they seem to have so little
understanding of software engineering that these criticisms can be made
at all is frightening.  Of course, what can we expect:  its from
Microsoft, after all.

Comments appreciated.

/*************from the XQL mailing list
today**********************************/


> 2 Lack of compositionality
>
> Especially the extensions for grouping, sequence, and multidocuments
> have unclear semantics. They are and can be explained for simple
> expressions, but give rise to multiple possible interpretations
> when substituting simple XQL-subexpressions (typically a single
> element like "title") with more complex XQL-subexpressions (e.g.
> (author | title)). More details on this in the specific remarks.
>
> Thus, XQL with these extensions is not compositional; XQL expressions
> can not be arbitrarily built from other XQL expressions. For the time
> being, the only way to maintain these extensions is to syntactically
> restrict XQL such that multiple and inconsistent interpretations
> are avoided. However, this makes the language clumsy. Gerald Huck
> has prepared a BNF, which respects some of the necessary syntactic
> restrictions (I have only a printout, so I can't include it here).
>
> Incidentally, XPath and its relatives have similar problems.
> E.g. The expression book/(title | author) is not syntactically
> possible with XPath. Consequently, the grammar/language is
> unnecessarily clumsy.
>
> 3 Formalia
>
> The general structure of the document is a tutorial based on
> examples. That is ok. But some of the examples are not even
> described in terms of the results they deliver. That is not
> ok.
>
/*****************much more along the same lines left
out*****************************/

Comments greatly appreciated.

Best Wishes,

Ann Tecklenburg
email:  at@ingenuity-sw.com